Hierarchical Modelling of Aquatic Species Distributions

ABSTRACT Aim Species Distribution Models (SDMs) have traditionally been developed in a terrestrial context, and their application to aquatic ecosystems presents unique challenges. These include predicting species distributions across spatially constrained environments and incorporating environmental drivers that operate across nested spatial scales. We address these challenges by exploring various spatially‐explicit model training strategies and novel hierarchical global‐to‐regional model structures to assess the scaling of environmental filters across climatic (broad), hydroclimatic (regional), and hydromorphological (local) gradients. Our goal is to establish a method for modeling aquatic species distributions that accounts for the distinct biogeography of freshwater systems. Innovation We demonstrate the application of a systematic analytical approach to model aquatic species distributions by investigating three critical dimensions: (i) the optimal spatial extents of training for SDMs in freshwater ecosystems; (ii) the effect of predictor combinations across climatic, hydroclimatic and hydromorphological factors on predictive performance; and (iii) the interplay of these interacting drivers from global to regional scales. This comprehensive cross‐comparison forms the core of our workflow for assessing overall model robustness across diverse model settings for aquatic taxa. Main Conclusions We used Iberian freshwater fishes as a case study. Our results show that models constrained to a species' watershed of occurrence, meaning they are trained strictly within its boundaries, deliver high predictive accuracy as opposed to wide‐scale training areas. The h 5 basin level emerged as the most parsimonious and highest‐performing training extent across global, global‐to‐regional, and explicitly regional models, and any further reduction of the watershed area resulted in an overall decline in model performance. For widespread species, particularly for native ones, the global niche was found to be the most important variable in our hierarchical models, demonstrating that regional studies must account for niche truncation to avoid biased results. Furthermore, our results show that climate and hydroclimatic drivers outperform locally structured, hydromorphological‐based models. This indicates that despite the importance of site‐specific features, the distribution of freshwater fish is primarily governed by broad‐to‐regional environmental changes. Finally, our results provide insights into the scaling of the environmental filtering processes that dictate freshwater fish distributions. We indicate that climate provides a macro‐environmental baseline defining the broad potential realized niche, upon which hydroclimatic factors refine regional habitat suitability while hydromorphological variables introduce local habitat heterogeneity. Recognizing that these broad‐to‐regional climatic and hydroclimatic drivers possess the highest explanatory power in this multi‐scale environmental mosaic, our method provides a basis for standardizing SDMs in aquatic systems, with the rigorous protocol established serving as a foundational model for large scale applications.